tuitab

A fast, keyboard-driven terminal explorer for tabular data.

Open CSV · JSON · YAML · TOML · Parquet · Excel · SQLite · DuckDB straight from your shell — filter, sort, pivot, join, compute columns, and chart distributions without leaving the terminal.

tuitab data.csv                   # open a file
tuitab config.toml                # JSON / YAML / TOML open as a browsable tree
tuitab deploy.conf                # unknown extension? the contents decide
tuitab -t yaml weird.dat          # or force a format outright
tuitab orders.csv customers.csv   # browse several files as a list
cat data.csv | tuitab -t csv      # read from a pipe

New to tuitab? Jump to the Quick start, or read the full Documentation — available in English and Русский.

Highlights

• Tabular and structured formats — CSV/TSV (auto-delimiter), JSON, JSONL/NDJSON, YAML, TOML, Parquet, Excel (xlsx/xls), SQLite, DuckDB. Browse a whole directory, or pipe data in over stdin.
• Nested data, edited in place — JSON/YAML/TOML open as a table over the real document: Enter dives into a nested object or list, ( expands a nested column into one column per key, m switches between record and key/value layouts, and edits write back into the document, not into a flattened copy. Saving re-serialises the tree, so structure survives.
• Vim-style navigation — hjkl, gg/G, page jumps, sticky pinned columns.
• Instant analysis — per-column statistics, frequency tables, and charts (histogram, bar, line, grouped bar) rendered right in the terminal.
• Reshape on the fly — pivot tables, JOINs across files, transpose, group by, deduplication, computed columns and window functions (rank, running total, lag/lead, share of a group) from an expression language with and / or / not.
• Clean, fast, type-aware — Polars-backed engine, Everforest theme, undo/redo, currency / percentage / date column types.
• Export and convert — write back to CSV, TSV, Parquet, JSON, JSONL, YAML, TOML, Excel, or SQLite. Converting between structured formats is just a different extension: open config.toml, save as config.yaml. Yank rows to the clipboard as TSV, CSV, JSON, or Markdown.
• MCP server — tuitab --mcp exposes the same engine to an AI assistant, so it computes over your data instead of guessing at it.

See it in action

Charts — histogram, bar, line, grouped bar

Press V on any column. Numeric columns get a Freedman–Diaconis histogram; categorical columns get a frequency bar chart. Pin a date or category column first with ! to draw line charts and grouped bars.

Pivot tables

Pin the column(s) to group by, place the cursor on the column to spread across, press W, and type an aggregation formula such as sum(revenue).

JOIN across files

Press J for a step-by-step wizard: pick another file (or an open sheet), choose INNER / LEFT / RIGHT / OUTER, and select the key columns.

Computed columns

Press = and type an expression. Arithmetic, string and date functions, and conditionals are all supported — the new column appears right next to the cursor.

Installation

Cargo (crates.io)

cargo install tuitab

Installs three commands: tuitab, plus the shorter aliases ttab and ttb.

Homebrew (macOS / Linux)

brew tap denisotree/tuitab
brew install tuitab

Arch Linux (AUR)

yay -S tuitab          # pre-built binary: tuitab-bin
# or build from source:
git clone https://aur.archlinux.org/tuitab.git && cd tuitab && makepkg -si

Debian / Ubuntu

Download the .deb from the Releases page, then:

sudo dpkg -i tuitab_*_amd64.deb

Pre-built binaries

Grab a tarball for your platform (Linux / macOS, x86_64 / aarch64) from the Releases page.

cargo build --release

The default build bundles DuckDB and SQLite from source, so the binary is fully self-contained — no system libraries required. This compiles DuckDB's C++ core (~5 min the first time).

To skip that and link a system DuckDB instead:

brew install duckdb                # macOS
sudo apt install libduckdb-dev     # Debian / Ubuntu
cargo build --release --no-default-features

Quick start

cargo install tuitab
tuitab data.csv

• Move with h j k l; jump with gg / G.
• Sort the current column: [ ascending, ] descending, r to reset; z[ / z] add a second key.
• Chart it: V. Per-column stats: I. Frequency table: F. Group by pinned columns: gb. Rank, running total or a group share: zw.
• Select rows by an expression: |!=amount > 1000.
• Add a column: = then e.g. revenue / units.
• Save / export: Ctrl+S. Quit: q. Help at any time: ?.

Usage

tuitab [OPTIONS] [FILES]...

Arguments:
  [FILES]...  One or more files to open, a directory, or '-' for stdin.
              Pass multiple files to browse them as a list.
              Defaults to the current directory.

Options:
  -d, --delimiter <CHAR>   Column delimiter (auto-detected if omitted)
  -t, --type <FORMAT>      Format when reading from stdin: csv, tsv, txt, json
      --mcp                Run as an MCP server on stdio (see below)
  -h, --help               Print help
  -V, --version            Print version

Browse several files

tuitab orders.csv customers.csv products.parquet

A directory-style listing opens with each file as a row. Press Enter to open one; Esc or q to go back.

Pipe mode

psql -c "SELECT * FROM orders" --csv | tuitab -t csv
sqlite3 app.db ".mode csv" ".headers on" "SELECT * FROM users" | tuitab -t csv

Stdin accepts csv, tsv, txt, json, jsonl, yaml, and toml. For Parquet/Excel/SQLite, open the file directly.

MCP server — let an AI assistant use the engine

tuitab --mcp speaks the Model Context Protocol over stdio. An assistant handed a data file can then compute over it with tuitab instead of doing the arithmetic in its head:

claude mcp add tuitab -- tuitab --mcp

Four tools: tuitab_inspect (columns, types, row count, sample rows), tuitab_query (fifteen operations composed as a pipeline — filter, group by, window functions, pivot, join, dedup and the rest — returning JSON or writing xlsx/csv/parquet), tuitab_describe (per-column statistics), and tuitab_jq (jq programs over nested JSON/YAML/TOML).

There is no SQL and no arbitrary code: the model sends structured operations, each mapping onto a function tuitab already had, and gets back numbers Polars computed. The server documents itself — the tool list and usage notes are sent to the model on connect.

Every operation is shared with a keybinding rather than written twice, and a test holds the two surfaces together: adding an action to the terminal fails to compile until someone says how a model reaches it.

Costs no extra dependencies: the protocol is newline-delimited JSON-RPC, which serde_json already covers.

Full details in the MCP server guide.

Keybindings

The essentials — see the full keybinding reference for every command (column ops, clipboard, dedup, and more).

Non-QWERTY layouts (ЙЦУКЕН, QWERTZ, AZERTY) are transparently remapped, so the hotkeys work regardless of your keyboard.

Documentation

Full guides, organised by topic, in two languages:

Acknowledgements

tuitab is inspired by VisiData — a brilliant terminal spreadsheet multitool by Saul Pwanson. If you find tuitab useful, check out VisiData too. Built with ratatui, Polars, and crossterm.

Contributing

Bug reports, feature requests, and pull requests are welcome. See CONTRIBUTING.md.

License

Apache-2.0 — see LICENSE.